I have no direct experience, so you can ignore the rest of this if you want. What's the cost to development from “slowing down” due to integration/acceptance t

I have no direct experience, so you can ignore the rest of this if you want. What's the cost to development from “slowing down” due to integration/acceptance tests before deployment vs. stopping development entirely to go back and fix errors that were deployed (including potential real dollar costs to customers being affected by the error)? If you had to choose between slowing down development by requiring tests vs. accruing technical debt by deploying potentially buggy code without the tests as a gate, which makes economic sense?